Joint state estimation and prediction in noisy wireless sensor networks

State estimation and prediction is an important problem in wireless sensor networks. It has strong influence on many aspects of sensor network systems, such as (i) fault tolerance, fault detection, and availability of the whole system after partial failures; (ii) system security in terms of confidentiality, integrity, and availability; (iii) resource consumption and energy efficiency; (iv) scalability and maintainability; and (v) estimation error reduction. In this paper, we present a joint state estimation and prediction scheme for Correlated Noisy Wireless Sensor Networks (CNWSN). In our scheme, we made two real assumptions about wireless sensor network model. First, the observed sensor data include both underlying process noise and measurement noise. Second, the actual sensor data values (i.e. without noise) are correlated with each other. The correlation between sensors might be either already known from the knowledge about underlying process from the system blueprint or could be estimated during monitoring. We propose the Kalman Filtering (KF) based approach to estimate and predict the states for correlated sensors in the network. Experimental results demonstrate that our KF-based join state estimation and prediction approach produces more precise estimation by exploring the inherent correlation among sensor data.